wapr.tugon.ph: A Secure Helpline for Detecting Psychosocial Aid from Reports of Unlawful Killings in the Philippines

Availability and affordance of information communications technology has provided additional medium to monitor human rights violation. Reporting, extraction, collection and verification of reports through natural language processing and machine learning techniques can now be integrated into one system. The processed narratives becomes readily available for response, monitoring, interventions, policy-making and even as evidence in court. This paper discusses the design and development of wapr.tugon.ph, a block-chain enabled NLP-based platform that provides a simple yet effective way of reporting, validating and securing human rights violation reports from victims or witnesses. wapr.tugon.ph allows for SMS-based and web-based reporting of human rights violation. Reports are processed for detection of emotions using NRCLex, and behaviors using Stanford Parser and modified Multi-Liason algorithm from narratives which serve as input to assess wellness. A total of 5,418 records were obtained from Reddit’s subreddits and HappyDB corpus to serve as baseline corpora for our model. Our best psychosocial wellness detection model produced an accuracy and F1 score of 84% on validation set (\(n = 1,426\)) and 87% on test set (\(n = 666\)). An ethereum private blockchain is implemented to record all transactions made in the system for authenticity tracking. Findings underscore the importance of providing a system that assists in determining the appropriate psychosocial intervention to victims, families and witnesses of human rights violation. Specifically, the study contributes a framework in embedding a combined sentiment and behavior model that outputs: sentiments that are used to assess mental wellness, behaviors that are used to assess physical needs, and detection of wellness that serves as input to refer victims, families of victims and witnesses to appropriate agencies.